Upload
others
View
4
Download
0
Embed Size (px)
Citation preview
This is a repository copy of Mapping hazard from urban non-point pollution: A screening model to support sustainable urban drainage planning.
White Rose Research Online URL for this paper:http://eprints.whiterose.ac.uk/2539/
Article:
Mitchell, G. (2005) Mapping hazard from urban non-point pollution: A screening model to support sustainable urban drainage planning. Journal of Environmental Management, 74 (1). pp. 1-9. ISSN 0301-4797
https://doi.org/10.1016/j.jenvman.2004.08.002
[email protected]://eprints.whiterose.ac.uk/
Reuse
See Attached
Takedown
If you consider content in White Rose Research Online to be in breach of UK law, please notify us by emailing [email protected] including the URL of the record and the reason for the withdrawal request.
White Rose Research Online
http://eprints.whiterose.ac.uk/
Institute of Transport StudiesUniversity of Leeds
This is an author produced version of a paper published in the Journal of Environmental Management. This paper has been peer-reviewed but does not include final publisher formatting or journal pagination. White Rose Repository URL for this paper: http://eprints.whiterose.ac.uk/2539/
Published paper Mitchell, G. (2005) Mapping hazard from urban non-point pollution: A screening model to support sustainable urban drainage planning - Journal of Environmental Management, 74(1) 1-9
White Rose Consortium ePrints Repository [email protected]
Mapping hazard from urban non-point pollution: A screening
model to support sustainable urban drainage planning
Gordon Mitchell.
The School of Geography, The University of Leeds, Leeds, LS2 9JT, UK.
Tel: +44-113-343-6721, Fax +44-113-343-3308, E-mail: [email protected]
Abstract
Non-point sources of pollution are difficult to identify and control, and are one of the
main reasons that urban rivers fail to reach the water quality objectives set for them.
Whilst sustainable drainage systems (SuDS) are available to help combat this diffuse
pollution, they are mostly installed in areas of new urban development. However, SuDS
must also be installed in existing built areas if diffuse loadings are to be reduced.
Advice on where best to locate SuDS within existing built areas is limited, hence a
semi-distributed stochastic GIS-model was developed to map small-area basin-wide
loadings of 18 key stormwater pollutants. Load maps are combined with information on
surface water quality objectives to permit mapping of diffuse pollution hazard to
beneficial uses of receiving waters. The model thus aids SuDS planning and strategic
management of urban diffuse pollution. The identification of diffuse emission ‘hotspots’
within a water quality objectives framework is consistent with the ‘combined’ (risk
assessment) approach to pollution control advocated by the EU Water Framework
Directive.
Keywords: Stormwater; Diffuse pollution; Sustainable Drainage System;
River basin planning.
1
1 Introduction
Pollution from urban non-point sources is one of the main reasons that rivers fail to
reach their water quality objectives. In Scotland, for example, 20 % of all water quality
failures are attributed to urban non-point sources (SEPA, 1996). Sustainable drainage
systems (SuDS) able to reduce diffuse pollution are available, including structural
source controls (e.g. swales, filter drains, permeable surfaces, wetlands) and 'good
housekeeping' best management practices, with advice on their implementation now
available in the form of design manuals (CIRIA, 2000). Understandably, SuDS practice
has focussed on new urban developments, as these offer opportunities to consider SuDS
at the earliest planning and design stages. However, the majority of diffuse urban loads
draining to surface waters come from existing developments; hence consideration must
be given to retro-fitting these areas with SuDS if diffuse pollution impacts are to be
reduced.
Advice is available on identifying sites in existing built areas where SuDS
implementation is feasible (Makropoulos et al., 1998) or acceptable (Ellis, 1998), the
latter considering criteria such as groundwater quality objectives. However, whilst it has
long been known that diffuse urban loadings can vary by an order of magnitude or more
between ‘clean’ and ‘dirty’ urban catchments (Ellis, 1991), little effort has been made to
identify sites that are significant in terms of load and potential impact on receiving
waters, and hence which deserve further detailed assessment and possible SuDS
implementation. This omission is significant within the context of the EU Water
Framework Directive (2000/60/EC), which requires that all emissions to water be
identified, quantified and managed.
2
This paper reports on the development of a nonpoint source preliminary planning and
assessment model designed for application to urbanised river basins, particularly those
in NW Europe and the UK. The model is consistent with the EU Water Framework
Directive’s combined approach to pollution that seeks a reduction of emissions within
limits defined by water quality objectives, and which operates at the river basin scale.
The model can: (a) quantify diffuse urban emissions, and map the location of areas
producing the poorest stormwater quality runoff within a river basin; (b) identify areas
which present the greatest hazard to the beneficial uses of receiving waters; and (c)
assess the impact of land use change on diffuse loads. Thus the model can be used to
identify urban areas where desired development presents the least risk to runoff quality,
and more importantly can assist in prioritising existing built sites for further detailed
assessment with a view to possible SuDS implementation.
2 Pollutants Addressed
Pollutants to be addressed in the model were selected using a simple risk assessment,
consistent with the approach subsequently recommended in the EU Water Framework
Directive for identifying pollutants of "management concern". Firstly, evidence of
widespread contamination was assessed through literature review, including evidence
from the US nation-wide urban runoff programme (NURP) (US EPA, 1983). Second,
intrinsic hazard was assessed through a review of pollutant toxicity, and the frequency
with which urban stormwater fails ecological and human health standards, based on the
NURP priority pollutant programme (Cole et al., 1984). Failure of water quality
standards in UK urban rivers provided further evidence of possible contamination. The
resultant provisional list was further refined through consultation with prospective end
3
users in the UK (including the Environment Agency, Scottish Environmental Protection
Agency, Water Research Centre, Water Utilities, and local authority urban drainage
practitioners), and following definition of the stormwater quality modelling approach.
The pollutant parameters addressed are: total suspended sediment, BOD, COD, total
metals (cadmium, chromium, copper, iron, lead, mercury, nickel and zinc); nutrients
(total nitrogen, total kjheldahl-nitrogen, total phosphorous, soluble phosphorous and
ammoniacal-N); and oil and grease.
3 Modelling Approach
There are numerous approaches to estimating urban stormwater loadings. These include
databases, regression analyses, unit load, sediment potency, volume-concentration and
build-up and wash-off methods (Marsalek, 1991). There is no universal 'best' approach,
and current advice is to match the modelling effort to the study objectives, avoiding the
use of large complex models when smaller simpler ones will suffice (FWR, 1994). This
study aimed to develop a preliminary planning and assessment model suitable for
identifying urban diffuse pollutant 'hot spots' within a river basin, for a wide range of
pollutants. Considering this objective, advice on model selection (Reckhow et al.,
1985), and a literature review of stormwater modelling (Mitchell et al., unpublished), it
was decided to develop a semi-distributed stochastic model with annual load modelled
as the product of runoff volume and a pollution concentration variable, the site mean
event mean concentration. The event mean concentration (EMC) is the total mass load
of a chemical yielded from a storm, divided by the total storm discharge. A storm EMC
4
is thus calculated as the area under the loading rate curve divided by the area under the
flow rate curve (or is simply taken as the flow weighted mean concentration in sampling
programmes). The mean value from multiple storms monitored at the same location is
the site mean EMC, and is the parameter used in the load assessments.
This modelling approach was adopted for several reasons. First, volume-concentration
methods often perform better than regression models (Brown, 1987; Pandit, 1997), and
give load estimates comparable to, and in some cases better than those of complex
build-up and wash-off models (Chandler, 1994; Charbeneau and Barrett, 1998). For
example, application of a volume-concentration model in the USA, showed that 77 % of
124 load estimates were within a factor of two of the estimates derived using
deterministic models (HSP-F and SWMM), and with 90 % certainty, were within 95 %
of observed loads, for catchments up to 2 000 km2 (Chandler, 1994). Such studies
indicate that volume-concentration methods, if applied with sensitivity to local
conditions, can provide load estimates of sufficient accuracy to inform SuDS planning,
and at a fraction of the cost of more complex models.
Secondly, observations from the UK (Mance and Harman, 1978; Moy et al., 2003),
USA (US EPA, 1983) and France (Hemain, 1986) have shown that the mean EMC of a
site is not correlated with annual runoff volume, simplifying the modelling exercise as
interactions between runoff volume and pollutant concentration could be ignored. Third,
many pollutants can be addressed, and new ones easily added as EMC data become
available. Fourth, EMC application allows uncertainty in loads estimation to be assessed
through probabilistic methods, and finally, basin scale applications are possible whist
5
maintaining a useful spatial unit of analysis. Once the approach was determined, site
mean EMC values and a runoff model appropriate to a UK application were required.
As land use is a determinant of both runoff and EMC, the model design proceeded
iteratively until a land use structure suited to both the runoff model and the EMC
database emerged.
Whilst the volume-concentration approach has been applied before, notably in the USA,
no urban diffuse pollution screening model (of any sort) suitable for basin scale
application in Europe is available. This paper presents the first such model. The study
applies the volume-concentration approach, applying a runoff model validated for use in
Europe, and derives European (and UK) specific EMC values via a review of 678
monitored urban catchments. The hazard mapping capability (see 6 below) also permits
a simple assessment of diffuse loading hazard to receiving waters. The model thus
supports management initiatives seeking to control emissions at source within a water
quality standards framework, whilst also providing a basin scale assessment, both
requirements of the Water Framework Directive.
4 Model Construction
4.1 Runoff Model
The requirements of the runoff model are: (a) an ability to quantify annual runoff
volume (event runoff estimation was unnecessary); (b) applicability at the basin scale
with a spatial unit of analysis <500m2; (c) sensitivity to spatial variability in annual
rainfall; (d) response to change in land use or land cover; (e) applicability to the UK
6
study area; and (f) ability to be implemented at low cost within a desktop Geographic
Information System (GIS), making it suitable for screening level applications.
A review of urban runoff models showed that statistical rainfall-runoff methods were
most appropriate. From over a hundred such methods reported in the literature, the
runoff volume algorithm (RUNVOL) of the Wallingford Modified Rational Method
(DoE, 1981) was selected. This function has the form:
RUNVOL = ((0.829 PIMP + 25.0 SOIL + 0.078 UCWI -20.7) / 100) x P x A
Where: RUNVOL is annual runoff (mm)
PIMP is the coverage by impervious surfaces connected to a storm sewer (%).
SOIL is a unitless variable developed during the UK Flood Studies Report
research (NERC, 1975). The index is based on winter rainfall acceptance, a soil
hydrologic characteristic which is broadly infiltration potential and the reverse
of runoff potential. Soil surveyors allocate SOIL values considering
permeability, groundwater level, and slope.
UCWI (Urban Catchment Wetness Index), is antecedent wetness (mm).
P is annual rainfall (mm).
A is catchment area.
RUNVOL matches the requirements of the runoff model detailed above. It has been
derived from the largest UK urban runoff database, comprising 510 storm events from
17 catchments, and is used to quantify runoff volume inputs to complex deterministic
7
models (e.g. WASSP, WALLRUS) that are currently used in UK sewer design. Thus
the method is proven, with potential for application in a basin scale distributed model.
4.2 GIS database
To support annual runoff volume and load modelling, a GIS database was developed to
address:
(a) Impervious area (PIMP). This impermeability variable was estimated using land use
data (see b below), and corresponding land use-impermeability coefficients derived
from surveys of NW London, Bolton, Nottingham and Dundee (Ellis, 1986; Ellis,
pers. comm.), conducted as feasibility preparation for the MOSQUITO sewer
quality model. This approach was used as other methods such as map interpretation,
field or air photo survey are prohibitively costly at the basin scale, and hence not
suited to a screening model. From the surveys, typical impermeability values were
obtained for 12 urban land use and residential density classes. Values for motorways
were drawn from the CIRIA 142 report (Luker and Montague, 1994).
(b) Urban land use. Through the CORINE project the European Environment Agency
provides digital land use data for Europe, collected through remote sensing, air
photo interpretation and field verification (EEA, 2000). Data is classified in a
hierarchical structure, with 44 land uses (11 are urban) at the most resolved tier.
Data is available in vector format (the smallest polygon is 25 ha) and as a raster
image (the smallest unit is 25 m) based on the vector data. Using the raster data, a
8
map was developed to address the land use categories of the EMC and
impermeability databases. Main highway areas were delineated using OS Meridian
road centre line data and the standard design width by road type.
(c) Residential density data was required to permit appropriate selection of the
impermeability coefficient described in (a) above, and potentially a density
dependent EMC value. This data was drawn from the SURPOP database (Anon,
1999), an ESRC facility in which demographic census data has been subject to
surface population modelling, a standard demographic technique designed to
overcome the problem of false spatial representation of demographic data when
using administrative boundaries. From the SURPOP database, residential density
data is available on a 200 x 200 m grid for the entire UK.
(d) The rainfall acceptance potential variable SOIL was mapped by digitising the SOIL
map presented in the UK Flood Studies Report (NERC, 1975);
(e) Annual rainfall. Daily rainfall for up to forty years was derived for Meteorological
Office records for 600 sites in the study region. For each site, annual and return
period annual averages were determined, and subject to area weighted Theissen
polygon modelling so as to generate rainfall maps for the river basin;
(f) The UCWI data was derived using the Wallingford method recommended
procedure, which relates UCWI to annual average rainfall via a power function
9
(DoE, 1981). Using this function, and the appropriate rainfall map described above
(e), a UCWI map was generated for the basin.
(g) Catchment area. Determined by the grid cell size adopted in the GIS.
4.3 EMC database
Site mean EMC values are recommended by the US EPA for use in pollutant discharge
assessment, a local government requirement under the Clean Water Act. They form the
basis, for example, of Department of Commerce National Oceanic and Atmospheric
Administration assessments of non point source loadings to coastal waters, part of the
National Coastal Pollution Discharge Inventory. However, US values are potentially
inappropriate in a European or UK context due to differences in urban structure and
land use activities. EMC values for UK and northern European applications were
therefore derived through analysis of an EMC database, constructed through literature
review (Mitchell, unpublished). In total, 160 urban stormwater quality studies were
included in the database, addressing 678 monitored catchments. Of these, 242
catchments were in northern Europe (Denmark, Finland, France, Germany, Norway,
Russia, Sweden, Switzerland and the Netherlands); and 71 in the UK. The majority of
the non-European data was drawn from the USA, Japan, New Zealand and Australia.
The database was used to: (a) assess normality in site EMC data; (b) identify any
differences in site EMC by land use; and (c) recommend EMC values for use in UK and
northern European urban stormwater screening analyses.
10
Analysis demonstrated that, for most pollutants, the site mean EMC conforms very well
to a predictable (log-normal) distribution, as indicated by Q-Q plots (e.g. Figure 1) and
Shapiro-Wilk normality tests. This finding, not previously demonstrated for European
site mean EMC data, allows reliable selection of probabilistic and central tendency site
EMC values. Differences in site mean EMC by land use are a priori assumed to occur
in the literature, but land use specific EMC values are often used without statistical
justification. Difference tests show that significant differences in EMC do occur
between land use classes in the northern European data set (Pt-stat > 0.05), but not always
in the UK data (sample size was often prohibitively small) or in the global (all
countries) data, where confounding effects across continents are assumed to vary more
than within northern Europe, masking the influence of land use.
Figure 1 about here
Given these significant differences, site mean EMC values could be recommended for
most pollutants for different land uses (residential, industrial and commercial, multi-
lane highways, other main roads, urban open) or combinations of these land uses (e.g.
all roads). However, for several pollutants (Cd, Cr, Fe and Hg) a small sample size
meant that only a single EMC value addressing all urban land could be recommended
(Table 1). Clear differences in EMC were not found by residential density class, but
were found between commercial and industrial land uses. However, the latter categories
were not treated separately in the GIS-model as they could not be resolved in the
available land use data.
11
Table 1 about here
5 Model Implementation
5.1 Probabilistic Load Mapping
The diffuse pollution screening model was applied to the Aire basin, Yorkshire, an area
of 2 057 km2 drained by the Aire and Calder rivers and their 24 main tributaries. The
basin includes the urban centres of Leeds, Bradford, Huddersfield and Halifax, and has
a population of 1.9 million people. The model was developed using the GIS MapInfo©
with the grid algebra extension Vertical Mapper©, but could readily be implemented
with any GIS with a grid algebra capability. All data was converted to a grid format
with a default cell size of 200 x 200 m, or 10 x 10 m if roads were addressed.
For each pollutant, the GIS-model was applied to calculate runoff volume per grid cell,
which was then coupled with an EMC value appropriate to the land use, so as to
generate a pollutant unit area load (UAL) in kg ha-1yr-1. Figure 2 illustrates the annual
non-point source loading of copper for a part of the Aire basin. Observed data to test the
validity of these loadings is not available (a motivation behind the model development),
but results are within the range of observed UAL's reported in the literature for northern
Europe.
Figure 2 about here
12
Loadings under non-average conditions can also be determined. The EMC data is log-
normally distributed, so an EMC value can be determined for any probability of
occurrence. Similarly, from the rainfall records, it was possible to generate ten year
return period annual rainfall maps, in a similar manner to flood frequency analysis. As
runoff volume is independent of EMC, percentile values representative of non-average
conditions can be chosen for both parameters (or just one if desired), and applied to
assess the effect of extremes of pollutant concentration and runoff on diffuse loadings.
5.2 Land Use Change Impact
The model was run to assess the effect of land use change on diffuse loads. Calderdale
Council, one of four local authorities in the Aire basin provided digital 'before' and
'after' land use maps for their district. The 'after' map represents land use and residential
densities in the area, assuming all of the permitted developments detailed in the 2005-
2016 Unitary Development Plan are executed. This includes provision of 1 800 new
residential properties in the area, although many are on brownfield sites. The analysis
indicates, for example, that this level of development would contribute an additional
530 kg yr-1 of copper to the river Calder. This is a relatively modest amount in the
context of a basin load of over 19 tonnes yr-1 (Neal, 2000), but indicates that anticipated
land use change in the basin can be expected to significantly raise total diffuse inputs.
5.3 UAL Mapping
The distributed load maps, superimposed over an OS 1:50 000 scale map, are used to
identify sites that merit further investigation and possible SuDS implementation.
13
However, these maps have a spatial resolution that makes manual identification of 'hot
spots' difficult at the basin scale. To address this problem, area averaging techniques,
such as the moving window average, were considered, but it was decided to sum cell
based data by small catchment, as this would assist with the subsequent hazard mapping
phase. For the Aire basin, 840 sub-catchments were identified using Hydrologic
Modelling 1.1, an Arc View routine that uses digital elevation data to determine flow
paths for incident rainfall, supported by manual vector editing. Improved catchment
boundaries could be defined through application of Standing Technical Committee 25
sewer network data, but the process was considered adequate for preliminary planning
purposes. For each catchment, results were mapped as the UAL, and as percentile UAL
(e.g. the top 5 % of catchments in load terms).
6 Hazard Mapping
To enhance the utility of the GIS-model, a simple risk assessment was conducted
relating modelled annual load to environmental quality standards (EQSs) for receiving
waters. The Aire basin was subdivided into separate catchments for which observed
annual discharge data is available from undisturbed gauging stations. From the
catchment area and discharge data, and the required EQS (the permitted maximum
pollutant concentration), a maximum acceptable unit area load (UALMax) was
determined. In effect, the UALMax is a quantitative expression of the annual cumulative
environmental carrying capacity of the receiving waters, ignoring short term acute
effects. For each pollutant, hazard maps are then generated simply as UAL / UALMax for
each sub-catchment, giving a unitless hazard score.
14
Hazard mapping adds new information to the appraisal, as the UALMax variable
recognises spatial variability in the environmental quality objectives of receiving waters
(e.g. upstream waters have higher quality standards). The hazard of diffuse pollution to
alternative uses of receiving waters can also be mapped. Beneficial use classes (e.g.
Cyprinid fishery, Salmonid fishery, Potable water abstraction) are defined by water
quality standards for a combination of pollutants, where the range of pollutants and/or
the quality standard vary according to the use class. By averaging hazard maps for each
pollutant within a beneficial use class (using the appropriate EQO), it is possible to map
the hazard posed to that beneficial use from diffuse urban sources. As with the load
maps, the scores for hazard to beneficial use can also be expressed in percentile classes
to identify those diffuse source areas posing the greatest hazard to the relevant
beneficial use. Figure 3 shows, for the Aire basin, the hazard to the Salmonid beneficial
use class. For each parameter addressed by the model, and included in the Salmonid
use class classification, hazard maps were developed. These parameter specific hazard
maps (each with a unit less score) have then been averaged, to indicate hazard to the
Salmonid environmental quality objective.
Figure 3 about here
The hazard assessment is not a full risk assessment, as the model does not address short
term acute effects, and ignores factors of gully pot and in pollutant amelioration (or
enhancement) resulting from dilution or sedimentation. Exposure pathways, dilution in
the receiving waters, and impact recovery potential are also ignored. Addressing these
factors requires application of more complex (e.g. mixing) models, which is considered
15
incompatible with the preliminary planning objective of the hazard mapping model.
Rather, the results of the ‘hotspot’ and hazard mapping procedures are intended to assist
in identifying sites which particularly merit more detailed appraisal using site specific
procedures recommended in guidance such as that produced by CIRIA (2000).
Decisions over the most appropriate form of management (including application of
SuDS or other regulatory controls) and considering site constraints can then be made.
7 Retrofitting Cities with SuDS
New urban developments are increasingly being addressed by SuDS, and this is to be
welcomed. However, at best, this can only prevent further increase in non-point urban
loads. To address current water quality failures attributed to diffuse sources, it is
essential to implement SuDS in existing built areas. This is a major challenge, as
existing developments place physical and design constraints on SuDS implementation,
although advice to support this process is emerging (Ellis, 1998; Makropoulos et al.,
1998).
Where retrofitting urban areas with SuDS has been considered, evaluations have shown
that there are many opportunities to install SuDS, and that doing so can have positive
effects on the quality of urban runoff. Sieker and Klein (1998), for example, report that
surface water discharges in a large Berlin catchment could no longer be addressed using
conventional end of pipe treatment due to insufficient space for settling tanks,
unfavourable hydraulic conditions and a high cost to address the whole catchment.
SuDS, including a variety of infiltration systems, were considered as an alternative. For
the entire city of Berlin, it was concluded that a disconnection of 30 % of the
16
impervious area could be easily achieved using SuDS, and that the resulting reduction in
discharge made it possible to convert existing retention tanks to soil filter tanks, with a
much improved overall purification efficiency than with settling tanks alone.
8 Conclusions
Diffuse urban pollution is a significant problem requiring retrofitting of existing built
areas with sustainable urban drainage systems (SuDS). However, prioritising locations
for possible SuDS implementation is constrained by the difficulty of assessing diffuse
loadings on a basin scale. To address this problem, a GIS-model has been developed
which can be used in urban diffuse pollution appraisal and SuDS planning. The model
can: (a) map the location of diffuse pollution 'hotspots', under a range of probabilistic
conditions; (b) identify those areas which present the greatest hazard to beneficial uses
of receiving waters; and (c) assess the impact of land use change on non-point source
runoff quality. The model is intended to act as a screening tool to guide subsequent site
appraisal using more detailed modelling or monitoring. The requirements of the EU
Water Framework Directive are addressed by the model which facilitates pollution
appraisal at the river basin scale, aids investigative monitoring, and supports
management of emissions at source within a water quality standards framework.
Pollutant loads are quantified using a volume-concentration technique, which has been
shown to perform as well or better than more complex and costly deterministic models.
From a review of 160 stormwater quality studies reported in the published and grey
literature, site mean event mean concentration (EMC) coefficients for 18 urban nonpoint
17
source pollutants were identified for the UK and/or northern Europe. Where sample size
permitted, analysis of the EMC database revealed log-normality in both UK and
northern European stormwater quality data, allowing derivation of mean and percentile
EMC coefficients, and screening assessments using probabilistic modelling. In addition,
statistically significant differences in pollutant site mean EMC by land use were
identified for northern European data, introducing greater spatial heterogeneity into
urban diffuse pollution screening assessments.
Finally, the recommended EMC coefficients are applicable to any northern European
runoff volume-pollutant concentration assessment. However, the model is particularly
well suited to a UK application, as it uses an established and widely accepted UK runoff
algorithm. The model can be implemented using readily available, low cost data, and
any desktop GIS with a grid algebra capability.
Acknowledgements
I am indebted to colleagues Jim Lockyer and Adrian McDonald for their essential
contributions. Hugh Duncan (CRCCH, Australia), Brian Ellis, Mike Harford (CMBC),
Thorkild Hvivted-Jacobsen (University of Aalborg), Sandy Elliot (NIWA), the
Environment Agency, SEPA, WRc, and Yorkshire Water all provided valuable
assistance. The research was funded by the EPSRC. OS data © Crown Copyright 2001
(licence 024674).
18
References
Anon, 1999. SURPOP V2.0. MIMAS (http://www.census.ac.uk/cdu/surpop).
Brown, R.G., 1987. Comparison of five methods used to calculate annual or seasonal
surface water loading. In: Nix, S. and Black, P. E. (Eds.), Symposium on monitoring,
modelling and mediating water quality, 303-312.
Chandler, R.D., 1994. Estimating annual urban nonpoint pollutant loads. Journal of
Management in Engineering, 10 (6), 50-59.
Charbeneau, R. J. and Barrett, M.E., 1998. Evaluation of methods for estimating
stormwater pollutant loads. Water Environment Research, 70 (7), 1295-1302.
CIRIA, 2000. Sustainable urban drainage systems: Design manual for Scotland and
Northern Ireland, C521, CIRIA, London.
Cole, R.H., Frederick, R. E., Healey, R. P. and Rolan, R.G., 1984. Preliminary findings
of the priority pollutant monitoring project of the nation-wide urban runoff program.
Journal of the Water Pollution Control Federation, 56 (7), 898-908.
DoE., 1981. Design and analysis of urban storm drainage: The Wallingford procedure.
Volume 1. Principles, methods and practice. National Water Council Standing
Technical Committee report, 28, Department of the Environment, London.
EEA, 2000. CORINE: Co-ordination of information on the environment. Technical
guide. European Environment Agency, Copenhagen.
19
Ellis, J.B., 1986. Pollutional aspects of urban runoff. In: Torno, H. C., Marsalek, J., and
Desbordes, M. (Eds.). Urban runoff pollution, Springer-Verlag, Berlin, 1-38.
Ellis, B.J., 1991. Urban runoff quality in the UK: Problems, prospects and procedures.
Applied Geography, 11, 187-200.
Ellis, J.B., 1998. Source control for non-point urban runoff: A sustainable option for
stormwater quality management? In: Novotny, V. (Ed.), Proceedings of the 3rd annual
conference on diffuse pollution, IAWQ, Edinburgh.
FWR, 1994. Urban pollution management manual, FR/CL 0002, Foundation for Water
Research, Marlow, 41-61.
Hemain, J.C., 1986. Statistically based modelling of urban runoff quality. In: Torno, H.
C., Marsalek, J., and Desbordes, M. (Eds.). Urban runoff pollution, Springer-Verlag,
Heidelberg, 277-304.
Luker, M. and Montague, K., 1994. Control of pollution from highway drainage
discharges. Report 142, CIRIA, 151.
Makropoulos, C., Butler, D., and Maksimovic, C., 1998. A GIS based methodology for
the evaluation of suitability of urban areas for source control application. In: Novotny,
V. (Ed.). Proceedings of the 3rd annual conference on diffuse pollution, IAWQ,
Edinburgh.
20
Mance, G. and Harman, M., 1978. The quality of urban stormwater runoff. In:
Helliwell, P.R. (Ed.). Urban storm drainage, Proceedings of the international
conference, University of Southampton. Pentech Press, London, 603-617.
Marsalek, J., 1991. Pollutant loads in urban stormwater: Review of methods for
planning level estimates. Water Resources Bulletin, 27 (2), 283-291.
Moy, F., R. W. Crabtree, and Simms, T. (2003). Long Term Monitoring of Pollution
from Highway Runoff: Final Report & CD. Environment Agency P2-038/TR1. ISBN 1
84432 208 4
Neal, C., House, W. A., Leeks, G.J.L., Whitton, B. A., and Williams, R. J., 2000. The
water quality of UK rivers entering the North Sea. The Science of the Total
Environment, 251, 557-573.
NERC, 1975. Flood studies report. Natural Environment Research Council, London.
Pandit, A., 1997. Estimation of annual pollutant loads under wet weather conditions.
Journal of Hydrologic Engineering, October, 211-218.
Reckhow, K.H., Butcher, J.B., and Marin, C.M., 1985. Pollutant runoff models:
Selection and use in decision making, Water Resources Bulletin, 21 (2), 185-195.
Sieker, H., and Klein, M., 1998. Best management practices for stormwater-runoff with
alternative methods in a large urban catchment in Berlin, Germany. Water Science and
Technology, 38 (10), 91-97.
21
SEPA, 1996. State of the environment report 1996. Scottish Environmental Protection
Agency, Stirling, 22.
USEPA, 1983. Final report of the nation-wide urban runoff program, Volume 1, United
States Environmental Protection Agency, Washington D.C.
22
Figure 1. Log-normality of site mean EMC: Oxidised nitrogen for industrial and
commercial land use in northern European cities.
Figure 2. Total diffuse urban copper load (kg ha-1yr-1) in the Brighouse area of the Aire
basin, Yorkshire, UK.
Figure 3. Urban nonpoint source hazard to the Salmonid beneficial use class in the river
Aire basin, Yorkshire, UK.
Table 1. Site mean EMC values applied in the diffuse pollution screening model
23
Table 1. Site mean EMC values applied in the diffuse pollution screening model
Pollutant Land use a Mean 1st Quartile
3rd Quartile
Data source b
N
TSS mg l-1 UO 126.3 57.0 279.8 G 18 IC 50.4 18.1 140.4 NE 34 R 85.1 37.6 192.5 NE 37 MLH 194.5 110.1 343.5 NE 16 MH 156.9 62.2 396.3 NE 6
BOD mg l-1 UO 7.9 3.5 18.2 G 4 IC 9.9 5.9 16.7 NE 29 R 8.5 5.1 14.1 NE 27 H 23.9 17.5 32.6 NE 11
COD mg l-1 UO 36.0 20.0 64.6 G 16 IC 146.2 121.3 176.1 NE 6 R 80.0 53.2 120.4 NE 20 H 136.5 89.1 209.2 NE 9
Cd ug l-1 All 2.2 1.3 3.7 NE 39
Cr ug l-1 All 7.3 3.5 15.0 NE 19
Cu ug l-1 UO 27.9 19.8 39.2 G 6 DU 51.1 22.3 117.1 NE 44 H 80.3 43.2 149.5 NE 21
Fe mg l-1 All 2.98 1.42 6.28 G 77
Pb c ug l-1 UO 60.6 28.8 127.4 G 11 IC 132.6 55.8 315.4 NE 11 R 140.5 91.6 215.5 NE 24 MLH 330.1 197.7 551.1 NE 14 MH 201.0 107.7 375.0 NE 10
Hg ug l-1 All 0.27 0.10 0.74 G 25
Ni ug l-1 UO 14.8 10.2 21.6 G 2 DU 30.4 18.2 50.6 NE 13
Zn ug l-1 UO 203.0 102.0 403.9 All 8 IC 188.6 84.7 420.2 NE 13 R 296.9 192.8 457.2 NE 25 MLH 417.3 284.0 613.3 NE 14 MH 253.1 97.7 655.5 NE 10
24
Table 1. / Cont.
Pollutant Land use a Mean 1st Quartile
3rd Quartile
Data source b
N
Total P mg l-1 UO 0.22 0.08 0.58 G 21 IC 0.30 0.16 0.54 NE 6 R 0.41 0.24 0.72 NE 18 MLH 0.28 0.15 0.52 G 35 MH 0.34 0.17 0.67 G 21
Soluble P mg l-1 UO 0.056 0.018 0.174 G 9 IC 0.156 0.070 0.345 G 24 R 0.198 0.109 0.359 G 45 H 0.178 0.101 0.313 G 4
Total N mg l-1 UO 1.68 0.86 3.27 G 14 IC 1.52 0.89 2.60 G 54 R 2.85 1.73 4.71 G 119 H 2.37 1.52 3.71 G 26
TKN mg l-1 UO 1.21 0.73 2.02 G 11 IC 1.54 1.06 2.23 G 34 R 2.40 1.54 3.74 G 84 H 1.60 2.75 0.47 G 20
NO2+3 mg l-1 UO 0.84 0.43 1.65 G 8 IC 0.60 0.40 0.92 G 32 R 0.98 0.50 1.91 G 68 H 0.81 0.63 1.03 G 17
NH4-N mg l-1 UO 0.10 0.10 0.10 G 1 DU 0.56 0.30 1.06 UK 37
Oil & Grease mg l-1 UO 0.60 0.60 0.60 G 1 DU 4.24 1.21 14.89 NE 18
a Land use: IC is Industrial and Commercial; R is Residential; UO is Urban Open;
MLH is Multi-Lane Highway (Motorway); MH is Main Highway (excludes MLH);
H is Highway (MLH + MH); (DU is Developed Urban (IC + R + not stated/mixed
but excluding UO, MLH, MH and H); All is All urban land uses.
25
b Data regions: G is Global, using all countries in the database; NE is Northern
European countries only; UK is United Kingdom data only.
c Due to the historical reduction in use of leaded petrol, the first quartile value was
used to represent the mean EMC.
26
27
Fig 1
-3
-2
-1 -0 .8 -0 .6 -0 .4 -0 .2 0 0 .2 0 .4
L o g N O 2+N O 3 (m g l-1)
-1
0
1
2
3
No
rmal
Q
uan
tile
Fig 2
28
29
30
> 1.0
0.4 --0.6
0.6 - 0.8
0.2 - 0.4
< 0.2
Annual diffuse load: Total dissolved
copper (Kg ha-1 yr-1)
Fig 2 Key
Fig 3
31